Abstract
An acoustic voice evaluation system has been used for screening for laryngeal diseases by a vowel signal that has been recorded directly from a microphone (direct voice). In order to apply it to the voice signal recorded through the telephone line (telephone voice), careful examination must be made on whether or not essential degradation of the acoustic measures could take place through the telephone line in terms of screening.
Two hundred thirteen sustained vowel samples were extarcted from a set of 240 voice data, by excluding 27 vowels that showed no obvious vocal cord vibration. Three acoustic measures that have played important roles in the evaluation system for the direct voice were subjected to the acoustic analysis : (1) period perturbation quotient (PPQ), (2) amplitude perturbation quotient (APQ) and (3) normalized noise energy in a full band (NNEa)
The findings are : (1) The fundamental frequency (F0) does not change through the line, (2) strong correlation is observed on the three measures between the direct and telephone voices, resulting in the autocorrelation coefficients of 0.962, 0.797 and 0.865 for PPQ, APQ, and NNEa, respectively, (3) APQ and NNEa show rather larger values for the telephone voice than the direct voice, reflecting frequency characteristics of the line, and (4) a statistical measure on the separability of the three hoarsenessgrade groups (Wilks lambda) indicates no significant difference. From these findings, we conclude that the telephone voice can be used for the acoustic voice evaluation as well.
